Enhancing Hybrid Classification for Plant Diseases with Deep Feature Selection Based on Analytical Entropy and Statistical Method

This study presents an effective and efficient framework for plant disease classification using a combination of deep learning and traditional machine learning techniques. Among various evaluated models, ResNet-18 emerged as the most effective feature extractor due to its capability to capture rich, hierarchical features from leaf images. To enhance the relevance of the extracted features and reduce redundancy, an entropy-based feature selection method was employed, ensuring that only the most informative features were retained. For classification, the Support Vector Machine (SVM) outperformed other classifiers, demonstrating high accuracy and robustness in distinguishing between multiple disease categories. The classification system was tested across a range of plant disease classes, including Healthy, Rust, Leaf Spot, Mildew, and Blight, among others, depending on the specific dataset used. The synergy between deep feature extraction and entropy-guided selection significantly improved the overall performance of the classifier. This hybrid approach not only achieved precise disease identification but also ensured computational efficiency, making it suitable for real-world agricultural applications. The proposed framework holds promise for early disease detection and crop management, contributing to precision agriculture and food security by enabling timely and accurate intervention.

Keywords: Plant Disease Prediction, Deep Learning, Image Processing , classification , Machine learning and feature Extracion.